The field of robotic manipulation and exploration is moving towards more efficient and adaptive methods. Recent developments have focused on leveraging generative models to enhance information exploration in 3D environments and improve multimodal information processing. This has led to significant improvements in high-precision tasks, such as robotic manipulation, and has enabled more efficient generation of physically plausible actions. Additionally, researchers are exploring the use of diffusion-based policy learning methods and hierarchical Gaussian world models to improve multi-task bimanual manipulation and robust robotic exploration. Notable papers include: FlowRAM, which proposes a novel framework for region-aware perception and achieves state-of-the-art performance in robotic manipulation tasks. AnchorDP3, which presents a diffusion policy framework for dual-arm robotic manipulation that integrates affordance guided sparse diffusion policy and achieves a 98.7% average success rate in the RoboTwin benchmark. ManiGaussian++, which improves multi-task bimanual manipulation by digesting multi-body scene dynamics through a hierarchical Gaussian world model and significantly outperforms current state-of-the-art bimanual manipulation techniques.